An object detection apparatus that utilizes a radio wave such as a millimeter wave radar as a means for detecting the existence of a vehicle on a road may be used in order to, for example, grasp a traffic volume or ensure a traffic safety of the road. In such an object detection apparatus, it is generally difficult to output all of the processed data due to the limitation of communication capacity of a data output unit. Accordingly, data determined to be more useful need to be output first.
As for a vehicle-mounted object detection apparatus, an image processing apparatus has been known in which targets are selected and processed sequentially from a target nearest to an estimated route of an own vehicle predicted from the traveling state of the own vehicle or sequentially from a target nearest to the own vehicle as images having higher priorities in radar search results for an area located in front of the own vehicle. In this case, the radar mounted on the own vehicle detects an area of each object candidate (target) included in a captured image of the area located in front of the own vehicle to determine the priority of an image recognition processing. Accordingly, the image processing apparatus may select a target in a descending order of importance (degree of attention) and perform the image recognition processing on the selected target without performing the image recognition processing on images of all areas.
The image processing apparatus, when performing the image recognition processing on the target, for example, sets an image recognition range which is substantially a reversed trapezoidal shape, when viewed from the own vehicle, on a captured image and uses each area of which one half width or more is included in the image recognition range as an area which falls within the image recognition range to define only images of these areas as a target image for recognition. Accordingly, when selecting the object candidate from sequentially from a target nearest to the own vehicle located in front of the own vehicle based on the coordinate values, the object candidate which is not fallen within the image recognition range is excluded from the image recognition processing. See, for example, Japanese Laid-Open Patent Publication No. 2006-163879.
When detecting an object using a radar, many objects other than the target object for detection may be included in a detectable range. This corresponds to a case where many objects such as a road structure on a roadside are included in a short distance when detecting a vehicle on a road. It is undesirable that a vehicle detection radar selects unnecessary data such as the road structure first. However, desirable data may not be obtained in a method where the image recognition range is set as a substantially reversed trapezoidal shape when viewed from the radar (own vehicle) on the captured image and data is sorted in an ascending or a descending order of the radar output value based on the set image recognition range, for example, in a case where the radar is installed at the roadside. That is, data of, for example, the structures other than a target object for detection (e.g., a vehicle) may be frequently detected. Further, considering that the data processing capability of the radar is not so high, data may be sorted and extracted using a relatively simple processing.